Creative Logic Reconstruction Skill

An agent skill for reconstructing why a product, work, creator, team, company, or strategy is shaped the way it is.

It is written as a SKILL.md-based skill and is intended to work in agent runtimes that support this skill format, including Claude Code-style and Codex-style skill directories.

It is designed for outside-view analysis:

• Why did they make this choice?
• What scene, audience, constraint, or belief made this form rational?
• What tradeoffs did they accept?
• What can be learned, translated, or deliberately not copied?

The skill works for products, apps, agents, films, essays, books, videos, games, creators, teams, and startup strategy.

Why This Exists

Many analyses stop at summary or taste:

It has feature X.
It feels good.
The founder probably wanted Y.

This skill pushes the agent to reconstruct logic from evidence:

observed form
  -> target scene
  -> constraints and tradeoffs
  -> competing hypotheses
  -> confidence-calibrated reconstruction
  -> transferable lessons

It is not a psychological profiling tool. It should not diagnose real people, infer sensitive traits, or treat sparse evidence as proof of hidden motive.

Install

Copy the skill folder into your skills directory:

cp -R creative-logic-reconstruction ~/.agents/skills/

Use the skill root for your agent runtime. Common examples:

# Claude Code-style local skills
cp -R creative-logic-reconstruction ~/.claude/skills/

# Codex / agents-style local skills
cp -R creative-logic-reconstruction ~/.agents/skills/

# Other runtimes
cp -R creative-logic-reconstruction <your-skill-root>/

The important part is that the runtime can discover creative-logic-reconstruction/SKILL.md.

Usage

Example prompts:

Use $creative-logic-reconstruction to analyze why this app uses a system-level overlay instead of a main chat window.

Use $creative-logic-reconstruction to reconstruct the creative logic of this essay and what makes its structure work.

Use $creative-logic-reconstruction to analyze this startup's product wedge and what I should learn or avoid copying.

What It Adds

• Mode selection for product, work/content, creator/team, and strategy/investment analysis.
• Evidence discipline: observed, inferred, unverified, refuted.
• Misclassification checks to avoid being fooled by surface form.
• Contextual transfer without assuming the user's project.
• Behavioral reconstruction guardrails inspired by evidence-based inference, not profiling.

Output Style

The skill favors calibrated language:

• "The evidence supports..."
• "A plausible hypothesis is..."
• "This seems to optimize..."
• "Alternative explanation..."
• "Confidence..."

It avoids:

• certainty theater
• creator mind-reading
• gossip
• psychological diagnosis
• copying surface patterns without copying the context

License

MIT